Whisper small ap2 - Nuwan
This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.5955
- Wer Ortho: 34.8080
- Wer: 34.1567
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-06
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_steps: 50
- training_steps: 6000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
|---|---|---|---|---|---|
| 2.2182 | 0.1642 | 400 | 2.3026 | 80.3514 | 78.8883 |
| 0.8087 | 0.3284 | 800 | 0.8213 | 70.8486 | 69.5550 |
| 0.7561 | 0.4926 | 1200 | 0.7547 | 50.8240 | 49.9130 |
| 0.7118 | 0.6568 | 1600 | 0.7234 | 47.8373 | 47.0305 |
| 0.6374 | 0.8210 | 2000 | 0.7005 | 45.7117 | 44.8588 |
| 0.6521 | 0.9852 | 2400 | 0.6818 | 45.0158 | 44.1598 |
| 0.6017 | 1.1494 | 2800 | 0.6670 | 45.6405 | 44.8369 |
| 0.5899 | 1.3136 | 3200 | 0.6565 | 41.7684 | 41.0111 |
| 0.5954 | 1.4778 | 3600 | 0.6458 | 44.0289 | 43.3934 |
| 0.5406 | 1.6420 | 4000 | 0.6323 | 40.5343 | 39.7944 |
| 0.4964 | 1.8062 | 4400 | 0.6231 | 38.2753 | 37.6519 |
| 0.5695 | 1.9704 | 4800 | 0.6130 | 39.6655 | 38.9257 |
| 0.4507 | 2.1346 | 5200 | 0.6077 | 37.3778 | 36.6238 |
| 0.409 | 2.2989 | 5600 | 0.6014 | 34.6367 | 33.9242 |
| 0.4788 | 2.4631 | 6000 | 0.5955 | 34.8080 | 34.1567 |
Framework versions
- Transformers 4.57.1
- Pytorch 2.9.0+cu126
- Datasets 3.6.0
- Tokenizers 0.22.1